NBER WORKING PAPER SERIES
THE STRESS COST OF CHILDREN
Hielke BuddelmeyerDaniel S. Hamermesh
Mark Wooden
Working Paper 21223http://www.nber.org/papers/w21223
NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue
Cambridge, MA 02138May 2015
We thank Michael Burda, Luise Goerges, Matthias Krapf, and participants in seminars at a numberof universities and conferences for helpful comments. This study uses unit record data from the Household,Income and Labour Dynamics in Australia (HILDA) Survey and German Socio-Economic Panel (SOEP).The HILDA Survey project was initiated and is funded by the Australian Government Departmentof Social Services and is managed by the Melbourne Institute of Applied Economic and Social Research(at the University of Melbourne). The German data used in this publication are from the German Socio-EconomicPanel Study (SOEP) and made available to us by the German Institute for Economic Research (DIW),Berlin. No funding was received by any of the authors in support of this study. The views expressedherein are those of the authors and do not necessarily reflect the views of the National Bureau of EconomicResearch.
At least one co-author has disclosed a financial relationship of potential relevance for this research.Further information is available online at http://www.nber.org/papers/w21223.ack
NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies officialNBER publications.
© 2015 by Hielke Buddelmeyer, Daniel S. Hamermesh, and Mark Wooden. All rights reserved. Shortsections of text, not to exceed two paragraphs, may be quoted without explicit permission providedthat full credit, including © notice, is given to the source.
The Stress Cost of ChildrenHielke Buddelmeyer, Daniel S. Hamermesh, and Mark WoodenNBER Working Paper No. 21223May 2015JEL No. I31,J12,J13
ABSTRACT
We use longitudinal data describing couples in Australia from 2001-12 and Germany from 2002-12to examine how demographic events affect perceived time and financial stress. Consistent with theview of measures of stress as proxies for the Lagrangean multipliers in models of household production,we show that births increase time stress, especially among mothers, and that the effects last at leastseveral years. Births generally also raise financial stress slightly. The monetary equivalent of the costsof the extra time stress is very large. While the departure of a child from the home reduces parents’time stress, its negative impacts on the tightness of the time constraints are much smaller than the positiveimpacts of a birth.
Hielke BuddelmeyerUniversity of Melbourne111 Barry StreetCarlton, Melbourne, [email protected]
Daniel S. HamermeshDepartment of EconomicsRoyal Holloway University of LondonEgham, TW20 0EXUNITED KINGDOMand [email protected]
Mark WoodenUniversity of Melbourne111 Barry StreetCarlton, Melbourne, [email protected]
1
Insanity is inherited—we get it from our children. [Mark Twain]
I. Background
We ask whether the addition of a child to a family imposes costs that are not accounted for in
the immense literatures on the cost of children and on equivalence scales, and thus whether there are
hitherto unaccounted factors that affect the decision to have a child or that increase the perceived
costs of rearing a child. The literature on equivalence scales focuses solely on the monetary costs of
children (e.g., Muellbauer, 1977; Pollak and Wales, 1979; Bourguignon, 1999). The sparser literature
on the time costs of children (e.g., Gustafsson and Kjulin, 1994; Bradbury, 2008) engages in
accounting exercises, totalling up the amounts of time that each parent devotes to child care, and
perhaps valuing them, and examining gender differences and secular changes in time allocated to
child care.
Hamermesh and Lee (2007) constructed and estimated a model describing cross-section
differences in the extent of expressed time stress. The theoretical basis was Becker’s (1965) model of
the use of time and goods to produce commodities that contribute to a household’s utility. The
theoretical part of the study identified time stress as the Lagrangean multiplier on a household’s time
constraint and linked financial worries to the Lagrangean multiplier on its goods constraint. Using
cross-section data from Australia, Germany, Korea and the U.S., they found that individuals with
higher Beckerian full incomes expressed greater feelings of time stress, consistent with a more tightly
binding time constraint, and that they were less likely to express concerns about money (consistent
with a looser goods constraint).1
Our approach here combines these two strands of the literature: We examine the extent to
which people find that the time and goods constraints in their utility maximization bind more tightly
when a child is added to the household. We are not examining generalized responses to a birth, such
as happiness or life satisfaction (for the mixed results on these see, e.g., Stanca, 2012, Baetschmann et
al, 2012, Pedersen and Schmidt, 2014), nor are we examining emotional responses to particular
aspects of child-rearing (e.g., Connelly and Kimmel, 2013). Instead, we study how a specific life
1DeVoe and Pfeffer (2011) use several waves of the Australian data set to demonstrate the relationship between income and time stress.
2
event—the birth of a child—affects the empirical analogs of parameters that arise within a family’s
welfare maximization. We thus develop a new dimension of the cost of children; and, because
additional time loosens the time constraint while additional income loosens the goods constraint, our
approach allows us to extend the measurement of the monetary and time costs of children. We
complement the examination of the impact of births on the household’s utility maximization by
studying what might be viewed as the obverse of a birth—the departure of a child from the household.
To obtain these estimates we need data sets that contain respondents’ views of the time and
monetary stress that they perceive, our analogs to the Lagrangean multipliers in their utility
maximization. Longitudinal data are also required, since in order to identify the effect of an addition
to the household we need a household-specific baseline against which to compare the empirical
counterparts to the multipliers. Fortunately, since 2001 the Household, Income and Labour Dynamics
in Australia (HILDA) Survey has collected annual information from a panel of respondents on their
perceptions of time and financial stress. Also, since 2002 the German Socio-Economic Panel (SOEP)
has collected similar information biennially. We use both data sets in the empirical work here, thus
providing a check on the specific cognitive implications of the questions and on culture-specific
differences in couples’ responses to the birth of a child.
II. Theoretical Motivation and Considerations
Consider a household that combines goods (a vector xj) and the time of each spouse (vectors
TMj and TF
j) to produce a vector of commodities Zj (j=1, …, N) that determines its current utility:
(1) U = U(Z1(x1,TM1, TF
1), … , ZN(xN, TMN, TF
N)).
The maximization of this utility function, given the technologies of household production and the
household’s wage rates, WM and WF, unearned income I, and the vector of goods prices that it faces,
Pj, yields a utility-maximizing vector of demands for both time and goods inputs into the production
of each commodity.2
2Equation (1) describes current-period utility, but clearly a planned birth must, if parents are rational, raise lifetime utility. Thus a complete model would append a term like e-rTU(.), indicating the present value of the infinite stream of satisfaction from creating a dynasty. This extension rationalizes the possible increase in happiness engendered by children with the possible tightening of the time and goods constraints on which we focus. The utility function implies pooling of resources in household production. More complex assumptions
3
The demands for time and goods inputs are functions of these prices. Similarly, the
household’s Lagrangean multipliers on the spouses’ time, λM and λF, and on goods, μ, are functions of
the parameters facing the household — the wage rates, unearned income and goods prices. We can
thus write each as:
(2a) λMt = λM(WM
t, WFt, It, Pjt);
(2b) λFt = λF(WM
t, WFt, It, Pjt);
(2c) μt = μ(WMt, WF
t, It, Pjt),
where t is some time period. Comparing across households, we make the standard assumption that all
households face the same goods prices, so that these can be ignored here and in the empirical work.
The usefulness of the model comes from its prediction that higher W and I raise λM and λF and lower
μ.
We could estimate equations (2) directly from survey respondents’ answers on their perceived
time and financial pressures. Some individuals may, however, always feel pressured, and others may
feel less pressured, even in the face of the same objective circumstances. Also, the amount of pressure
generated by the birth may depend on its interaction with the family’s existing demographic structure.
Taking these considerations together, recognizing that all the information affecting maximization in
the previous period will be subsumed by the outcomes in that period, and linearizing (2), we can
rewrite the model as:
(3a) λMt = a1λM
t-1 + a2λFt-1 + a3μt-1 + α1WM
t + α2WFt +α3It + α4ΔKt + νM
t,
(3b) λFt = b1λM
t-1 + b2λFt-1 + b3μt-1 + β1WM
t + β2WFt +β3It + β4ΔKt + νF
t,
(3c) μt = c1λMt-1 + c2λF
t-1 + c3μt-1 + γ1WMt + γ2WF
t + γ3tIt + γ4ΔKt + ηt,
where the a, b and c are parameters describing the autoregressions, η and the ν are normally
distributed error terms, and ΔK, the focus of most of this study, denotes the change in the family’s
demographic structure, including crucially the addition of a child.
A potentially important issue here is the problem of the endogeneity of births in a year in
response to stress (both time and financial) in that same year. To model this potential endogeneity in
would not yield any additional readily testable implications about time or financial stress in the context of the data available to us.
4
this context, let us assume that, along with many other things described by the vector of variables X,
both expected time stress and expected financial stress affect the probability of having a child. Let S*
be the upper limit to perceived time stress (S) beyond which people will decide not to have a child,
and let F* be the analogous upper limit to perceived financial stress (F). Then assuming that the
couple has complete control over its fertility, the probability that a child is born is the joint
probability:
(4) Pr{ΔKi,t+1=1} = Pr{[αE(ΔXi,t+1) + βSit + εit < S*], [γE(ΔXi,t+1) + δFit + θit < F*]},
where ε and θ are normally distributed and presumably are not independent, and α, β, γ and δ are
parameters describing this probability for couple i. Equation (4) can be rewritten as the bivariate
probit:
(5) Pr{ΔKi,t+1=1} = Pr{[εit < S*- αE(ΔXi,t+1) - βSit ], [θit < F* - γE(ΔX′i,t+1) - δFit]}.3
There are several ways of dealing with this potential endogeneity. We could expand beyond
estimating (3a) - (3c) jointly to estimating them jointly with the selection equation (5). The difficulty
with this approach lies in finding exclusion restrictions appropriate for the four equations (the
couple’s financial stress, the time stress of each spouse, and fertility). An alternative approach would
argue that any biases to the estimates of the impact of a birth on time and financial stress that are
caused by the potential endogeneity of births will be negative. Those parents who expect smaller
increases in stress are those who are more likely to have a child. Thus we would expect that any
estimated positive impacts of a birth on stress that we find will understate the “treatment effect” that
would be observed if births were distributed randomly across the population of couples arrayed by the
impact of births and changing stress. As estimates of the local average treatment effect of a birth, our
estimates will be biased toward zero.
III. Data and Descriptive Statistics
Both surveys that we use provide nationally representative longitudinal data sets describing
the populations of the countries studied. The HILDA Survey asks the following question of survey
3Since having children is hardly an uncommon event, one might wonder why couples do not forecast its impact on their time and financial stress—essentially, why they might lack rational expectations about the effects of a birth on stress. Odermatt and Stutzer (2014) provide some evidence and arguments for why people do not forecast the impact of life events on a loosely related concept, their happiness, very well.
5
participants: “How often do you feel rushed or pressed for time?” with possible answers “almost
always,” “often,” “sometimes,” “rarely” and “never”. Thus we can index t = 2001, 2002, …, 2012,
which, allowing for lagged values, enables us to estimate autoregressions based explicitly on (3a) and
(3b) for eleven years of births. Participants are also asked to rate their satisfaction with their financial
situation on an eleven-point (0 to 10) scale ranging from ”totally dissatisfied” to “totally satisfied”,
allowing us to estimate autoregressions based explicitly on (3b). To provide comparability with the
scale on time stress, we collapse the responses to this latter question into five categories.4 Thus the
autoregressions that we estimate track the Lagrangean multipliers λ and μ. Since both spouses express
satisfaction with their financial situation, we estimate separate equations for each and test for the
equality of their responses to a birth. Couples are included until separation or a spouse’s death, and a
person who “re-couples” is reintroduced into the sample with the new partner if observed in two
consecutive years.
We currently have six waves of data from the SOEP with the necessary information, t = 2002,
2004, …, 2012, allowing, with the required lag, for five biennia of births. Biennially the SOEP has
included the question: “Think about the last four weeks. How often during this period did it happen
that you felt rushed or under time pressure?” with possible responses “always”, “often”, “sometimes”,
“almost never” and “never.” 5 Perhaps because of the differences in phrasing in the SOEP or in how
the answers are elicited, the distribution of responses to this question is tilted more heavily toward
being less rushed for time than in the HILDA Survey.6 The SOEP asks all respondents the same
question about financial stress as the HILDA Survey, and we treat responses exactly the same.7 Thus,
4We recode responses 0-2 as 5 (4.9 percent of the sample), 3-4 as 4 (9.3 percent), 5-6 as 3 (28.2 percent), 7-8 as 2 (45.2 percent), and 9-10 as 1 (12.4 percent). Here and throughout this study we weight all sample observations by their sampling weights. 5The SOEP uses a four-week reference period and employs a multi-mode approach, with data collected by both interviewer and via self-administration, whereas in the HILDA Survey this question is always administered as part of a separate self-completion questionnaire. 6In the SOEP the distribution (never to always) is 5.8 percent, 15.1 percent, 39.2 percent, 33.9 percent, and 6.0 percent. In the HILDA Survey the comparable distribution is 0.8 percent, 10.3 percent, 40.1 percent, 36.0 percent, and 12.8 percent. 7The percentages of observations in the five recoded categories in the SOEP are 6.2 percent, 13.7 percent, 27.4 percent, 40.8 percent, and 11.9 percent, remarkably similar to the distribution of responses to this question in the HILDA Survey.
6
except for relying on biennial observations, the estimates of the determinants of the analogs of λ and μ
are based on similar questions in the two data sets. Replacing the one-year by a two-year lag in (3a) –
(3c), we can estimate equations for Germany that resemble those for Australia very closely.8
Table 1 presents the statistics describing the couples included in the sub-samples from the
HILDA Survey and SOEP over which we estimate (3a) - (3c). Here and in all subsequent tables
involving the examination of the impacts of births we exclude couples in which the wife is over age
45. In the HILDA Survey sub-sample wives report being significantly more stressed for time than
their husbands (paralleling the greater time stress perceived by women generally that was reported in
Hamermesh and Lee, 2007), but both spouses feel roughly the same financial stress. Ten percent of
the couples produced a child between successive interviews (and thus between responses on time and
financial stress); and the majority had other children present too. Half the respondents reported being
in excellent or very good health, with a higher fraction of wives reporting this. During the average
week the husbands spent 46 hours working (in paid employment) and commuting, while their wives
spent nearly 24 hours per week in these market-related activities. Time spent in household production
was almost reversed, so that reported (not from time diaries) total market and non-market work time
was not quite identical for the spouses (see Burda et al, 2013).9 Average total annual earnings (in
2012 dollars) of couples were around A$96,000, while average unearned income (in 2012 dollars)
among these couples was about A$20,000.10
The descriptive statistics from the SOEP show quite similar patterns on time stress. Wives are
significantly more stressed for time than their husbands. Husbands, however, express significantly
more financial stress than their wives. About one-eighth of the couples experience a birth during a
biennium over the time period 2002-12 (implying, consistent with data on vital statistics, a lower 8We use PanelWhiz (Hahn and Haisken-deNew, 2013) to create the sub-samples that underlie all our calculations. 9The measure of household production constructed from the HILDA Survey data is the amount of time in a typical week spent on household errands, housework, outdoor tasks, caring for children (including the children of other people, if unpaid) and caring for disabled or elderly relatives. In contrast, the SOEP only allowed us to include time spent on a typical weekday. The list of activities, however, was similar, and included running errands, housework, child care, helping other persons in need of care, repairs to the house/car, and garden work. For further details, see the Data Appendix. 10In 2007, the mid-point of the sample, the Australian dollar was worth about $US 0.79. We deflated all monetary measures by the Australian CPI.
7
annual birth rate than in Australia). In line with popular perception, husbands report more market
work time than their wives, and wives report significantly more home production time on weekdays.
Average annual earnings of the couples are roughly €53,000 per year (in 2012 prices), which is
consistent with published data, but average unearned income, at about €7,100 per year, may be low
(although these are prime-age intact couples).11
IV. Preliminary Examination of Patterns of Stress
We initially estimate equations (3a) - (3c) separately for each spouse including a number of
controls. As a first step toward this, and to obtain a picture of how a birth/adoption alters the time and
goods constraints, we examine transitions of the empirical counterparts of λ and μ. Consider columns
(1) and (3) of the top panel of Table 2, which show the fractions of the samples for which time stress
increased, remained the same or decreased between annual interviews in the HILDA Survey sub-
sample, separately by gender and by the indicator for the addition of a child to the household.12
Husbands in households adding a child are more likely than other husbands to feel increasingly
stressed for time. Comparing the changes in time stress for men in the HILDA Survey yields a test
statistic of χ2(2) = 15.99 (p < .001). Wives’ time stress is increased even more significantly on average
by a birth: The same test for Australian women in this table yields χ2(2) = 24.97 (p < .001).
Columns (2) and (4) in the upper panel of Table 2 show the same changes (over two-year
periods) calculated for the couples in the SOEP. For men the results look quite similar, and the
trivariate distributions (more, the same, or less time stress) are only barely distinguishable (χ2(2) =
3.96, p = .14). Among wives, however, the patterns differ greatly, with a much greater fraction
exhibiting increases in time stress if a birth has occurred in the biennium (χ2(2) = 8.17, p = .02, on the
trivariate distributions).
In columns (1) and (3) of the bottom panel of Table 2 we present the analogous patterns of
changes in perceived financial stress from the HILDA Survey, again separately for husbands and
wives by the indicator for the addition of a child to the household. As with time stress, adding a child
increases financial stress for both spouses. Also as with time stress, perceived financial stress 11In 2007 the euro was worth about $1.34. All monetary measures are deflated by the German CPI. 12The results are even clearer in the full 5x5 transition matrices.
8
increases more among new mothers than new fathers. Comparing households without and with a birth
in the HILDA Survey, husbands in the latter group are more likely to perceive an increase in financial
stress than those in the former group (χ2(2) = 25.55, p < .001), but the difference between the changes
in financial stress among wives is larger and even more significant statistically (χ2(2) = 37.68, p <
.001).
Columns (2) and (4) in the bottom panel of Table 2 present the same calculations for biennial
transitions in financial stress from the SOEP. For both spouses there are more increases in financial
stress among those couples that experience a birth. Among men we cannot reject the hypothesis that
the trivariate distributions are the same (χ2(2) = 4.18, p = .12). For their wives, however, the
difference in the distributions is highly statistically significant (χ2(2) = 11.18, p = .004).
We can expand upon these one- or two-year transitions by examining averages of time and
financial stress for each year before and after a birth, thus accounting for any changes in stress that
might be missing from the models that include only one year of lags (but excluding the vector X, and
not based on comparisons to couples without a birth in a particular year or biennium). Figure 1
presents these measures for both husbands and wives in couples that produced a child, from four years
before the birth through four years after, in the HILDA Survey. The picture is of clear increases in
both types of stress for both spouses after a birth; but paralleling the results for Australia in Table 2,
the graph suggests that the increases are greater for the wife than for her husband and greater for time
than for financial stress. Indeed, the wife’s time stress continues to rise steadily each year after the
birth, while her financial stress remains constant. The husband’s time and financial stress both
diminish, although they remain higher than they were on average before the birth.
The patterns in the figure suggest care in interpreting the parameter estimates of (3a)-(3c). For
women, but not for men, there is a clear “Ashenfelter dip” in both time and financial stress in the year
before the birth, especially so for time stress (Ashenfelter, 1978). Indeed, perhaps the temporary
decrease in stress increases the couple’s interest in having a child, as the discussion surrounding
equations (4) and (5) suggests. Regardless, these findings indicate that estimates of the determinants
of current stress that include only one lagged value may overstate the impact of the birth for women in
9
the Australian data. For men there is no pre-birth dip in time stress, but financial stress is much lower
in the pre-birth year.
In the SOEP, for which the patterns of time and financial stress before and after a birth are
shown in Figure 2, there is no evidence of dips in either time or financial stress in the biennium before
a birth. There may in fact be no dips, but perhaps our inability to detect any could be due to the
relative infrequency with which the data on time and financial stress are collected.
V. Estimates of Models of Stress
Table 3 lists least-squares estimates of analogs to (3a)-(3c) using the HILDA Survey (again,
with separate estimates of the impacts on financial stress for husbands and wives). We include and
report on the impacts of each spouse’s time allocation, weekly earnings (and thus, since work hours
are included, implicitly the full prices of their time), the family’s unearned income, and the
respondent’s self-reported health. (See the Data Appendix for further details of these and the other
variables included.13) More time spent at market work or in household production increases time
stress for each spouse, with market work being especially stressful. (Given a fixed time budget, this
means that shifting away from leisure or personal time increases time pressure.) A higher hourly wage
appears to have no impact on time stress in these estimates, but among women, who do most of a
household’s purchasing, having a higher-earning husband or greater unearned income increases time
stress, providing some support for the idea that households combine time and goods. For both spouses
being in good health reduces both time and financial stress, presumably by adding to the efficiency of
household production.14
The birth of a child significantly increases the perceived time stress of both husbands and
wives. The impact, however, is three times greater on the wife’s time stress than on her husband’s,
confirming the evidence from the changes in time stress shown in Table 2. Independent of the wife’s 13Also included are vectors of indicators of the number and ages of other children in the household (0-4, excluding the newborn, 5-10, 11-15, 16-18), the respondent’s and spouse’s ages (31-40, 41+), and year indicators. We also include, as per the theoretical motivation, lagged values of the other three stress measures (e.g., in the case of husband’s time stress, the wife’s time stress and both spouses’ financial stress). In addition, we estimated each model using an ordered probit, with no qualitative difference from the least-squares estimates reported in Table 3. All four estimated impacts of a birth on stress are positive and statistically significant, and the average derivatives differed by less than 0.02 from the OLS estimates. The impact on the wife’s time stress is over twice that on her husband’s, while the impacts on the spouses’ financial stress are nearly identical. 14Some direct evidence supporting this assertion is provided by Podor and Halliday (2012).
10
greater shift from leisure/personal time to household production that raises her time pressure when a
child is born (since the equation held the allocation of time constant), the very fact of the birth has a
much larger effect on the time pressure that she perceives than on her husband’s.
The changes in Table 2 suggested that both husbands and wives perceive additional financial
stress with a birth, and holding constant for time allocation and full incomes this conclusion remains,
although neither effect is strongly significant statistically. The theoretical motivation in Section II
suggested that the spouses’ views of their financial stress might respond identically to a birth. Jointly
estimating the equations describing their perceived financial stress, we cannot reject the hypothesis
that the responses are equal (t = 0.15). The main conclusion here is that a birth causes increases in
both spouses’ perceptions of financial stress, with perhaps an insignificantly larger response by the
wife than the husband.
It is well known that women’s time in the market and in home production responds to a birth
(by decreasing and increasing respectively), so that the impacts of time use on stress are quite likely in
part generated by the birth itself. To circumvent what is essentially a problem of spurious correlation,
we re-estimate the models in Table 3 without the time-use variables. The impacts of a birth on
husbands’ time stress and both spouses’ financial stress are essentially unaffected by this deletion.
The parameter estimate on wives’ time stress drops from +0.254 to +0.214, an insignificant decline
and one that still leaves the wife’s response significantly above the husband’s. If we drop all controls
except the lagged values of the stress measures, the indicators of the spouses’ ages, and the year
indicators, the estimated impacts of a birth on the husband’s (wife’s) time stress become +0.060
(+0.136), and on their financial stress +0.079 (+0.152). The overall conclusion is that relatively little
of the impact of the birth on stress works through a re-allocation of time. Most is inherent in the
changed circumstances in the nature of the household’s combination of goods and time that are
generated by the addition of a child, circumstances that increase the wife’s time stress and probably
her financial stress more than her husband’s.
Table 4 presents the same estimates for the SOEP sample. Unsurprisingly, given the biennial
data here, the sizes of the impacts of lagged stress are smaller and of lower statistical significance than
those in Table 3. More important, while the birth has a large and significant positive impact on the
11
wife’s time stress, unlike in the HILDA Survey its impact on her husband’s time stress is not
statistically significant. Neither spouse’s financial stress is significantly affected by the birth,
however, and both impacts are tiny.15
Here too, given their weekly earnings an extra hour of market work in a week raises both
spouses’ perceived time stress; but while it has significant negative effects on the husband’s perceived
financial stress, it has no impact on the wife’s. Consistent with the role of the husband as the major
earner in most couples, his financial stress is barely affected when his wife works more, while hers
decreases substantially when her husband works more (at the same hourly earnings). Additional time
spent in home production raises the wife’s time stress. Given each spouse’s time use, when either
spouse earns more per hour (has a higher full income) the time stress of each spouse increases,
although not statistically significantly; and unsurprisingly each spouse’s financial stress diminishes
significantly. The one set of surprising results in Table 4 is the negative (albeit not statistically
significant) impact of additional unearned income on time stress, and its positive impact on financial
stress. As in the HILDA Survey, being in good health reduces both time and financial stress.
Excluding the time-use measures hardly alters the estimated parameters on the indicator of a
birth in the equations describing time stress nor in those describing financial stress. In the former the
estimate for men rises slightly to +0.073, while for women it falls slightly to +0.196. The estimates
for this indicator in the financial stress equation both remain tiny and statistically insignificant.
Deleting all the controls except the indicators for year and for respondents’ ages, the impact on men’s
time stress changes little (+0.050), while that for women remains statistically significant but falls
dramatically (+0.086).16
The amount of stress felt by new parents may be greater among first-time parents than others.
To examine this possibility we add an indicator for first births to all the equations. In the equations
15Ordered probit estimates of the four specifications reported in Table 4 yield similar results. The impacts of a birth on each spouse’s financial stress are statistically insignificant, as is the impact on the husband’s time stress, while the effect on the wife’s time stress is highly significant and positive. As in the HILDA Survey estimates, the average derivatives differed very slightly from the OLS estimates. 16Alternatively, one can expand the time measures by adding indicators for zero work hours for each spouse. Doing so very slightly increases the estimated impacts of a birth on time stress for each spouse in both data sets, and very slightly decreases the estimated impacts on financial stress.
12
describing time stress in the HILDA Survey the coefficients on this indicator were -0.010 (s.e. =
0.038) and -0.004 (s.e. = 0.040) for men and women respectively. In the equations describing
financial stress their counterparts were -0.063 (s.e. = 0.038) and 0.031 (s.e. = 0.041). In the SOEP the
extra impacts of a first child on time stress were -0.043 (s.e. = 0.049) for husbands and -0.052 (s.e. =
0.056) for wives. For financial stress the additional impacts were 0.037 (s.e. = 0.051) and -0.013 (s.e.
= 0.051). A fair conclusion is that there is no evidence that a first child adds more to time or financial
stress than do subsequent children.
As an extension to these basic estimates we examine whether the changes in time and
financial stress occasioned by a birth depend on the presence of older children in the household. We
thus interact the birth indicator with the vector of indicators for older children and re-estimate the time
and financial stress models for husbands and wives. In the HILDA Survey these interactions (four in
each model) are not statistically significant as a group or individually in describing time stress, but the
impacts on both husbands’ and wives’ expressed financial stress are significantly affected by the
presence of other children. Having a primary-school age child reduces the perceived financial stress
occasioned by a birth, while having a teenager raises it. In the SOEP the presence of older children
does not interact significantly with a birth to influence financial stress; but when a child under age 5 is
present, a birth increases the time stress that the mother feels after a birth. Taken together, the
estimates make it clear that the magnitudes of the effect of a birth on time stress do not vary much
with the ages or numbers of older children present in the household.
As noted earlier, one spouse’s idiosyncratic responses to a birth may interact with the other’s,
and each spouse’s perceived time pressure may be related to his or her perceived financial stress.
Since the equations include all the same variables, the only issue here is the extent to which the errors
in the four equations are correlated. In both samples, once we account for the X variables, the four
lagged measures of stress and the birth indicator, the only significant correlations are between the
spouses’ financial stress (r = +0.28 in the Australian data, r = +0.36 in the German data) and between
their time stress in the SOEP (r = +0.19).17
17These conclusions do not change qualitatively if we exclude the time-use measures or, indeed, all the other controls from the basic equations.
13
The presence of the pre-birth “Ashenfelter dip” in expressed time stress, especially among
wives, could be at least partly responsible for the estimated impacts of a birth on time stress. One way
to circumvent this problem is to estimate the models without any lagged measures of time stress, but
including person fixed effects. The estimated impact of a birth then becomes the difference between
the stress measure immediately after a birth and its person-specific average over the entire panel,
adjusted for current measures of time use, earnings and unearned income, health and family structure.
Estimating these fixed-effects models for Australia yields an impact of a birth on husbands’ time
stress of +0.113 (s.e. = 0.026), and on wives’ of +0.260 (s.e. = 0.028). For Germany the analogous
fixed-effects estimates are +0.068 (s.e. = 0.034) for husbands and +0.246 (s.e. = 0.034) for wives.
These estimated impacts differ little from those shown in Tables 3a and 3b. The results also differ
little if we estimate fixed-effects ordered probit models.
A potential difficulty with using fixed-effects estimation is that the impact of a birth on time
stress may remain high for several years after the birth, as is indeed shown in Figure 1. An alternative
approach to handling the dip (at the cost of shortening the sample period and losing observations) is to
use longer lags in the stress measures, so that the comparisons are to earlier expressions of stress
rather than merely to the previous year’s (or in the SOEP, the previous biennium’s). Re-estimating the
models in Table 3 by adding two- and three-year lagged measures of stress, the estimated impact of a
birth on husbands’ time stress increases to +0.134 (s.e.=0.044), while that on wives’ falls to +0.153
(s.e. = 0.049). In the SOEP we add lagged measures of stress from the interview four years before the
year after the birth, with the resulting estimated impacts of the birth on time stress equalling +0.039
(s.e.= 0.043) among husbands, and +0.176 (s.e.=0.044) among wives.
These two methods to account for the drop in perceived time stress during the year ending
before the decision to have the child yield somewhat different results. The overall conclusion,
however, is that the implied significantly positive impact of the birth on time stress is robust, and that
this effect remains greater on the wife’s time stress than the husband’s.
Does the effect of a birth on time and financial pressure increase or diminish over time? In
other words, are the effects that we have demonstrated temporary and caused by the birth, or do they
represent the persistent stress costs of a child? To answer this question for Australia we estimate the
14
same models as presented in Table 3, except that we include lagged terms for successively two, three
and four years in the birth indicator and in the stress measures. We restrict the sample to couples that
had no additional birth, so that we are examining how a birth between Years t and t+1 affects stress at
Years t+1 (the results in Table 3), t+2, t+3 and t+4. All estimates include the same other current-
period controls that were included in the specifications underlying the results in Table 3.
The estimates are reported in the top part of Table 5, measured in standard-deviation units of
stress. While the estimated effects on time stress fluctuate from year to year, with generally smaller
effects the more distant in the past the birth is, they remain positive, larger among wives than
husbands, and statistically significant among wives. The initial effects on financial stress diminish and
are essentially zero two years after the birth. The general conclusion here is that, at least for the four
post-birth years that the sample size allows us to follow these couples, time stress, especially the
wife’s, remains above what it was before the birth, while the extra financial stress essentially
disappears.
With the biennial data in the SOEP the specification of the lag structure must be different,
since taking more than two lags would remove most of the sample observations. Accordingly, in the
bottom row of the bottom panel of Table 5 we report the estimated (in standard-deviation units)
impacts of a birth between Years t and t+2 on stress at Year t+4, including lagged stress measures
from Year t and all the current-period controls. The upper row in this panel converts the estimates
from Table 4 into standard-deviation units. Between two and four years after the birth none of the
effects on stress are statistically significant; the wife’s time stress remains, however, substantially
positively affected, and both spouses’ financial stress is higher than before the birth.
Not surprisingly there are some major differences in the results between the two data sets.
Partly they occur because of the different frequencies at which stress is measured; and we can
examine the extent to which the difference in the frequency of the data on stress is generating the
different results by aggregating births in the HILDA over two years and re-estimating the models
describing current time and financial stress, using the same controls and a two-year lag in time stress.
Given this temporal aggregation, we lose nearly half the observations (but none of the births), as we
are only using observations from 2004, 2006, …, 2012. The results of estimating the models using
15
this aggregation look somewhat like those reported in Table 3, although the coefficient on births
describing women’s time stress is somewhat reduced (but remains statistically significant). The
difference in the frequency of the questions on stress between the two panel data sets explains some of
the differences in the results across the two countries/data sets but far from all.
The results may also differ because the questions eliciting time stress and the measures of
time inputs differ across the surveys. We account for those discrepancies by including an indicator of
whether the stress measures in the SOEP are elicited by an interviewer or are responses to a self-
administered questionnaire. Those respondents who were interviewed express significantly less stress
on both dimensions; but their time and financial stress respond to a birth almost identically to that of
respondents who completed a questionnaire.18 Finally, the results may differ due to differences in
child care and family leave policies between the two countries.
There is a remarkably consistent pattern throughout the results: A birth generates initial time
stress in the new mother, and that stress persists for at least four years. Moreover, it is greater than the
new father’s additional time stress, which in any case does not persist. There is much less evidence of
an increase in perceived financial stress felt by either spouse.19
VI. The Monetary Equivalent of the Time Stress
Since the largest immediate effect of a birth is on the time stress felt by new mothers, in
attempting to monetize the costs of stress we concentrate on that particular form of stress. While we
propose three approaches to calculating the monetary equivalent of the additional time stress felt by
mothers that is generated by a birth, there are undoubtedly many other simulations beyond those
examined here that might be proposed. But at least these three do give an indication of the magnitude
of the monetary amounts that are equivalent to the psychological burden of the birth.
18Another set of possible causes of the differences involves different policies on child care and family subsidies. While with two observations we cannot examine these possibilities, we did consider how an increase in the generosity of child payments in Germany after 2007 might have affected the estimates. Perhaps because of the resulting small sample sizes, or perhaps because it actually had no effect, when we disaggregate the SOEP sample into pre- and post-2007, we find no differences in the estimated impacts of a birth on time stress. 19Our findings are captured in a letter from a mother of two pre-school children (July 5, 2002, from Hannah Ebin): “With the kids and the house, I often feel I have four hours of tasks and only two hours to do them in.”
16
In all of the simulations we ask the question: what is the monetary transfer or infusion of
earnings that would reduce the new mother’s financial stress by an amount equal to the increased time
stress generated by the birth? We are not advocating these transfers; rather, we are using them as a
way of measuring the monetary equivalent of the mother’s time stress produced by the new child. The
measures of subjective stress (time and financial) are not directly commensurate, so we calculate all
effects in standard-deviation units. We conduct simulations to answer three questions:
Simulation 1: What transfer of weekly earnings from the husband to the wife would
reduce her financial stress by the same amount that the birth has increased her time
stress?
Simulation 2: What increase in the wife’s weekly earnings would decrease her
financial stress by the same amount that the birth has increased her time stress?
Simulation 3: What increase in the husband’s weekly earnings would decrease the
wife’s financial stress by the same amount that the birth increased her time stress? 20
We perform all three simulations for both the HILDA Survey and SOEP using the estimates in Tables
3. In addition to calculating these one-time transfers/infusions immediately after a birth, we also
calculate their cost per married couple if each couple, regardless of whether it experiences a birth in
the year (biennium in the SOEP), were to pay taxes annually into a fund to finance the transfers.
We show the results of these simulations in Table 6. The effects are remarkably large,
especially in the first simulation, where even in the HILDA Survey the required one-time transfer is
over twice the average husband’s annual earnings (and even the annual transfer from all couples
would exceed 20 percent of husbands’ annual earnings). Clearly, there is no reasonable transfer of
earnings from husband to wife that can compensate for the increased time stress that she experiences
with the new child. The other possible changes also suggest extremely large monetary equivalents of
the mothers’ time stress. Thus even the least costly (Simulation 2, and Simulation 3 in the SOEP)
implies payments during the first year of each child’s life whose annual cost to every couple (the few
20The difference among Simulations 1, 2 and 3 is that under Simulation 1 total household earnings remain unchanged, whereas in the others they increase.
17
new parents and all other couples) of over US$4000 per year would represent a substantial increase in
the burden of taxes/transfers.
These simulations suggest that the psychological cost of a new child is huge in comparison to
the monetary cost and, even more so, to the value of time that the new mother and father expend on
the addition to the family. While other simulations would generate different monetary comparisons to
the time stress experienced by new mothers, given our estimates it is doubtful that any reasonable
simulation would suggest that these costs are small. One might think that providing subsidized early
childhood care would reduce time stress; but a comparison of the coefficients in Tables 3 to the means
in Table 1 indicates that, even with no time spent in household production (including child care), a
birth generates substantial additional time stress for the wife.
VII. Experimenting with the Endogeneity of a Birth
While we have argued that selectivity into child-bearing will bias downward our estimates of
the impact of a birth on time and financial stress, we cannot demonstrate that proposition empirically.
It is a sensible theoretical assertion about behavior. Our estimates would thus be even more
convincing if we could find a satisfactory instrument for birth. Regrettably, neither of the data sets has
any other variables that one could not easily argue also affect time and/or financial stress directly, and
other variables that might predict birth (age, number of children of various ages, spouses’ earnings,
and time allocation) are also predictors of time/financial stress (and are included in (3a) – (3c)). The
finding of a pre-birth dip in women’s time stress, however, might make the dip itself an appropriate
instrument to identify a five-equation model of this process (describing each spouse’s time and
financial stress and also the birth).
The pre-birth drop in women’s time stress may be behavioral. As implied in (5), unusually
low time and financial stress should induce couples to select into the population of new parents. There
is also biomedical evidence that women with low stress, as measured by low values of a particular
biological marker, are more fecund (Louis et al, 2011). While we cannot distinguish the behavioral
from the biological in either of our data sets, the two effects work in the same direction.
Using the HILDA Survey we estimated an equation describing the probability of a birth that
included the lagged change in each spouse’s time and financial stress, plus the lagged indicators of the
18
number of children in each of the four age categories.21 In a linear-probability model the parameter
estimates on the husband’s and wife’s lagged change in time stress are +0.0077 (s.e. = 0.0044) and -
0.0170 (s.e. = 0.0044); those on the husband’s and wife’s lagged change in financial stress are -0.0098
(s.e. = 0.0043) and -0.0012 (s.e. = 0.0042). These effects are small as well as being statistically
insignificant in some cases.
Observing stress only biennially in the SOEP makes that data set a weak candidate for
investigating this predictor; and Figure 2 showed that unsurprisingly the dip in women’s time stress
between time periods t-4 and t-2 was much smaller than the dip observed between t-2 and t-1 in
Australia. Nonetheless, we used the SOEP to estimate a linear model describing the probability of a
birth as a function of each spouse’s changes in time and financial stress between periods t-4 and t-2
(i.e., including two measures of lagged changes in stress). The estimated impacts on the probability of
a birth were all small and statistically insignificant, and were unexpectedly positive.
Regrettably in both data sets the predictive power of the lagged measures of stress is quite
weak: In Australia the adjusted R2 in predicting whether a birth occurs is only 0.050, while in the
SOEP it is 0.024. The lagged stress terms would be very weak instruments, so we do not go further
and use them to endogenize births. Nonetheless, the findings here are fascinating, suggesting in the
HILDA Survey that declines in the wife’s time stress and in her husband’s financial stress may help to
induce the couple to have a child.
VIII. Emptying the Nest
The theoretical motivation in Section II was based on the addition of a child and demonstrated
how that demographic change would cause the time and goods constraints facing the household to
bind more tightly. The reverse change, the departure of a child, should have the reverse effect: It
should decrease the tightness of the constraints and lower measures of their empirical analogs—
21The parameter estimates change minutely if we add each spouse’s earnings and the household’s unearned income to the specification.
19
perceived time and financial stress. To examine this potential asymmetry, we investigate whether the
reverse effects exist and are equal but of opposite sign to those demonstrated above.22
Because very few children depart their parents’ households when the mother is age 45 or less,
we expand both samples by removing the restrictions on the mother’s age. This expansion of the
sample changes the averages of the crucial outcomes substantially (compared to the averages shown
in Table 1), decreasing in all cases.23 In the Australian data the average time stress is 3.10 and 3.31 for
men and women respectively, while the average financial stress is 2.36 and 2.32. In the SOEP the
means of time stress are 2.65 and 2.84, and of financial stress are 2.61 and 2.52, for men and women
respectively.
In Table 7 we present statistics describing changes in husbands’ and wives’ time and financial
stress depending on whether a child departed the household that year (within two years after a
departure in the SOEP), thus listing the results in the same way as those for births shown in Table 2.
In seven of eight comparisons (husbands-wives, HILDA-SOEP, time and financial stress) those
people who had a child leave the household were more likely to experience a decrease in stress, and
less likely to experience an increase, than those who did not. The only exception is in the distributions
of changes in financial stress among wives in the HILDA Survey.
In general, the results mirror those shown in Table 2 for births: A departure generally reduces
stress. Comparing the results here to those in Table 2, however, shows that the differences in changes
in stress between those who do or do not experience the demographic event are much smaller for
departures of children than they were for births. Indeed, the trivariate distributions are not statistically
different from each other for time stress among men and financial stress among women in either the
Australian or the German data. While the differences in the impacts of births and departures on time
22As with the impact of a birth on a couple’s happiness, the impact of a child’s departure on happiness has also been examined (Krekel, 2013). 23Without this expansion of the sample sizes we would observe very few departures of children, and those few would be highly non-randomly selected. Changing the sample definition obviously alters the age mix of the respondents. Thus in the samples used earlier the average ages of wives in the HILDA Survey and the SOEP were 35 and 37 respectively. Removing the age restriction raises these respective averages to 48 and 52. Throughout this section we also exclude observations for years (biennia in the SOEP) in which a couple experienced a birth.
20
stress are more pronounced among wives, even there the magnitudes of the differences and their
statistical significance are far below those of their counterparts in Table 2.24
We can explore the dynamics of time stress around this demographic event, as we did for
births in Figures 1 and 2, by considering averages of time and financial stress +/- four years around a
child’s departure. The results are shown in Figures 3 and 4, constructed exactly as their analogs for
births. The first thing to note is that, unlike for births in the HILDA Survey, here we find no pre-event
dip in either time or financial stress. Rather, in both surveys and for both husbands and wives, time
stress appears to diminish more or less steadily from at least two years before a child departs the
household; and it continues decreasing in all cases for two years after. In both surveys, and for both
spouses, financial stress also decreases from at least two years before the event; but the decrease stops
or even reverses itself within two years after the departure.
Going still further, we estimate equations with specifications like those reported in Tables 3
and 4, except that here the variable of interest is the departure of a child. To save space, in Table 8 we
report only the least-squares estimates of the impacts of the departure on the measures of each
spouse’s time and financial stress. While in both surveys the wife’s time stress decreases with the
child’s departure, the decreases are small compared to the increases shown in Tables 3 and 4, and they
are not (quite) statistically significant.
While these results weakly corroborate the prediction that having a child leave the house
loosens time constraints, they suggest that the responses to what might seem like opposing events are
in fact asymmetric. Births tighten the constraints much more than departures loosen them. Moreover,
the results imply that, unlike births, departures are associated with a nearly steady diminution of time
stress both before and after the event, with generally similar effects on financial stress.
IX. Conclusions and Implications
Using data from longitudinal surveys for Australia and Germany, we have demonstrated that
a birth causes a rise in mothers’ time stress that is not dissipated over the first few years of her child’s
life. There is some evidence of a similar but smaller effect on fathers’ time stress; and we find some
24Restricting departures to those that result in an empty nest (where no children remain the household) does not alter the conclusion. The differences between those with and without a final departure remain small.
21
weak evidence that a birth increases spouses’ financial stress. This demonstration is not that births
affect such inchoate concepts as well-being or life satisfaction. Rather, by analogizing time stress to
the Lagrangean multiplier on each spouse’s time constraint, and financial stress to the multiplier on
the household’s goods constraint, the results are consistent with a model with households maximizing
their utility given their full income.
The magnitudes of the impacts of a birth on time stress are substantial, especially for a new
mother. Calculating the extra earnings that the mother would have to receive to reduce her financial
stress by as much as the birth increases her time stress (measured in comparable standard deviation
units) suggests that the monetary equivalent of the time stress of a birth on average is huge.
Demonstrating the magnitude of this additional cost of children might justify the subsidies to new
parents offered in many countries that might be viewed as partial attempts to offset these
nonmonetary, but measurable costs of having and raising children.
The results also provide evidence of the expected reverse pattern of responses to demographic
events, in that a child’s departure from the household generally reduces spouses’ time stress. But these
negative effects appear to be much smaller than the positive effects of a birth. Implicitly, the pleasure
of having children is sufficient to offset the implicit additional lifetime stress that they cause parents.
This is obvious; but the novelty here is the demonstration of the magnitudes and time paths of that
stress.
Because of the limitations of the data sets—and especially the relatively short duration of the
panels—our ability to examine the dynamic effects of births and of departures from the household on
time and financial stress within a general model of household production has been limited. While this
research suggests that having children generates a permanent lifetime increase in perceived stress, the
long-term effects of a birth on stress can only be analyzed with longer panels than are currently
available. That and linking the impacts of births on time and financial stress to spouses’ bargaining
behavior in the household remain potentially fruitful avenues for additional study.
22
REFERENCES
Orley Ashenfelter, “Estimating the Effect of Training Programs on Earnings”, Review of Economics and Statistics, 60 (Feb. 1978): 47-57.
Gregori Baetschmann, Kevin Staub and Raphael Studer, “Does the Stork Deliver Happiness?
Parenthood and Life Satisfaction”, Department of Economics, University of Zurich, Working Paper Series no. 94, October 2012.
Gary Becker, “A Theory of the Allocation of Time”, Economic Journal, 75 (Sept. 1965): 493-517. François Bourguignon, “The Cost of Children: May the Collective Approach to Household Behavior
Help?”, Journal of Population Economics, 12 (Nov. 1999): 503-21. Bruce Bradbury, “Time and the Cost of Children”, Review of Income and Wealth, 54 (Sept. 2008):
305-23. Michael Burda, Daniel Hamermesh and Philippe Weil, “Total Work and Gender: Facts and Possible
Explanations”, Journal of Population Economics, 26 (Jan. 2013): 239-61. Rachel Connelly and Jean Kimmel, “If You’re Happy and You Know It, Clap Your Hands: How Do
Mothers and Fathers Really Feel about Child Caregiving”, IZA Discussion Paper No. 7531, July 2013.
Sanford DeVoe and Jeffrey Pfeffer, “Time Is Tight: How Higher Economic Value of Time Increases
Feelings of Time Pressure, Journal of Applied Psychology, 96 (July 2011): 665-76. Bjorn Gustafsson and Urban Kjulin, “Time Use in Child Care and Housework and the Total Cost of
Children,” Journal of Population Economics, 7 (July 1994): 287-306. Markus Hahn and John Haisken-DeNew, “Panel Whiz and the Australian Longitudinal Data
Infrastructure in Economics”, Australian Economic Review, 46 (Sept. 2013): 379-86. Daniel Hamermesh and Jungmin Lee, “Stressed Out on Four Continents: Time Crunch or Yuppie
Kvetch?” Review of Economics and Statistics, 89 (May 2007): 374-83. Clinton Hayes and Nicole Watson, “HILDA Imputation Methods”, Melbourne Institute of Applied
Economic and Social Research, University of Melbourne, HILDA Technical Paper Series 2/09, December 2009.
Christian Krekel, “Home Alone: The Effects of Children Leaving and Returning Home on Parental
Well-Being,” Unpublished paper, German Institute for Economic Research (DIW), 2013. Germaine M. Louis, Kirsten J. Lum, Rajeshwari Sundaram, Zhen Chen, Sungduk Kim, Courtney D.
Lynch, Enrique F. Schisterman and Cecilia Pyper, “Stress Reduces Conception Probabilities Across the Fertile Window: Evidence in Support of Relaxation”, Fertility and Sterility, 95 (June 2011): 2184-9.
John Muellbauer, “Testing the Barten Model of Household Composition Effects and the Cost of
Children”, Economic Journal, 87 (Sept. 1977): 460-87. Reto Odermatt and Alois Stutzer, “(Mis-)Predicted Subjective Well-Being Following Life Events,”
Unpublished paper, University of Basel, Oct. 2014.
23
Peder Pedersen and Torben Schmidt, “Life Events and Subjective Well-Being: The Case of Having Children,” IZA Discussion Paper No. 8207, May 2014.
Melinda Podor and Timothy Halliday, “Health Status and the Allocation of Time,” Health Economics,
21 (May 2012): 514-27. Robert Pollak and Terence Wales, “Welfare Comparisons and Equivalence Scales”, American
Economic Association, Papers and Proceedings, 69 (May 1979): 216-21. Luca Stanca, “Suffer the Little Children: Measuring the Effects of Parenthood on Well-Being
Worldwide,” Journal of Economic Behavior and Organization, 81 (March 2012): 742-50.
24
Table 1. Descriptive Statistics: Couples (Means and Standard Deviations)
HILDA (N=7,376) SOEP (N=7,525)
Variable* Husband Wife Husband Wife
Time stress 3.41 3.59 3.14 3.25 (0.85) (0.87) (0.97) (0.95)
Financial stress 2.45 2.43 2.67 2.56 (0.98) (0.97) (1.05) (1.06)
Child born in year 0.10 0.12 / Born in last 2 years (0.29) (0.33)
Child 0-4 0.46 0.17 (0.68) (0.40)
Child 5-10 0.53 0.50 (0.78) (0.68)
Child11-15 0.34 0.37 (0.66) (0.62)
Child16-18 0.13 0.18 (0.38) (0.43)
Excellent or very good health 0.51 0.57 0.59 0.61 / very good or good health (0.50) (0.49) (0.49) (0.49)
Work and commute time 46.23 23.78 40.43 20.05 / Work time (17.98) (20.21) (15.87) (17.60)
Home production time per week 25.22 50.10 3.73 9.65 / Home production time per weekday (18.72) (34.24) (3.10) (7.10)
Earnings: (2012)A$ per week 1269 584 794 294 / Earnings: (2012) € per week (1005) (610) (598) (336)
Unearned income: (2012) A$ per week 384 147 / Unearned income: (2012) € per week (1098) (276)
*The first variable label describes the HILDA measure, the second the SOEP measure.
25
Table 2. Year-to-Year Transition Matrices on Stress, with or without Birth, HILDA 2001-12, SOEP 2002-12*
No Birth: HILDA
(N=11,203) SOEP (N=6,571)
HILDA (N=11,228)
SOEP (N=6,567)
Birth: (N=1172) (N=954) (N=1216) (N=958)
Time Stress
Change in Stress Men, No Birth Women, No Birth
Increase 22.2 29.1 22.2 28.7
Same 54.8 43.0 53.8 43.5
Decrease 23.0 27.9 24.0 27.8
Men, Birth Women, Birth
Increase 25.7 28.6 32.5 37.8
Same 55.6 46.0 49.5 38.2
Decrease 18.7 25.4 18.0 24.0
Financial Stress
Change in Stress Men, No Birth Women, No Birth
Increase 22.6 27.1 23.6 25.2
Same 53.2 49.3 50.0 48.3
Decrease 24.2 23.6 26.4 26.5
Men, Birth Women, Birth
Increase 28.3 27.9 31.0 34.3
Same 51.2 51.3 50.2 41.5
Decrease 20.5 20.8 18.8 24.2
*The numbers of observations differ slightly for men and women in each category because we condition on item non-response on the control variables used in subsequent regressions.
26
Table 3. LS Estimates of the Effects of a Birth on Stress, HILDA* (N = 7,376)
Time Stress (5 to 1) Financial Stress (5 to 1)
Independent Variable: HUSBAND WIFE HUSBAND WIFE
Lagged stress (own) 0.547 0.507 0.498 0.466
(0.013) (0.015) (0.017) (0.016)
Birth in past year 0.093 0.254 0.063 0.072
(0.032) (0.040) (0.041) (0.039)
Excellent or very good health -0.087 -0.113 -0.152 -0.152 (0.019) (0.018) (0.021) (0.024)
Work and commute time/week (own) 0.007 0.009 -0.003 -0.001
(0.001) (0.001) (0.001) (0.001)
Home production/week (own) 0.002 0.002 0.001 -0.0003
(0.001) (0.001) (0.001) (0.0004)
Earnings (own) -0.006 0.022 -0.118 -0.172
(0.010) (0.020) (0.012) (0.024)
Work and commute time/week (partner) -0.0001 -0.0001 0.0001 -0.001
(0.001) (0.0007) (0.0008) (0.0008)
Home production/week (partner) -0.0001 -0.0001 -0.0002 0.0008
(0.0004) (0.0007) (0.0005) (0.0006)
Earnings (partner) 0.007 0.021 -0.073 -0.070
(0.020) (0.011) (0.024) (0.013)
Unearned income/week 0.008 0.024 -0.047 -0.048
(0.007) (0.009) (0.012) (0.009)
R2 0.383 0.374 0.430 0.393
*Also includes all three other lagged stress measures, a vector of measures of numbers and ages of children, year indicators and indicators of the respondent’s and spouse’s decadal ages (31-40 and 41+). Robust standard errors clustered on person identifiers are reported here and in subsequent tables reporting coefficient estimates.
27
Table 4. LS Estimates of the Effects of a Birth on Stress, SOEP* (N = 7,525)
Time Stress (5 to 1) Financial Stress (5 to 1)
Independent Variable: HUSBAND WIFE HUSBAND WIFE
Lagged stress (own) 0.310 0.303 0.368 0.319
(0.018) (0.020) (0.024) (0.021)
Birth in past year 0.052 0.212 0.012 0.014
(0.051) (0.058) (0.051) (0.055)
Very good or good health -0.216 -0.225 -0.176 -0.209
(0.029) (0.034) (0.031) (0.031)
Work and commute time/week (own) 0.015 0.012 -0.006 0.0004
(0.001) (0.002) (0.001) (0.0015)
Home production/week (own) 0.003 0.017 0.009 0.0009
(0.006) (0.003) (0.006) (0.0035)
Earnings (own) 0.080 0.129 -0.302 -0.291
(0.026) (0.064) (0.046) (0.075)
Work and commute time/week (partner) -0.001 0.0003 -0.001 -0.004
(0.001) (0.0012) (0.001) (0.001)
Home production/week (partner) 0.0005 -0.012 0.004 0.002
(0.003) (0.008) (0.003) (0.006)
Earnings (partner) 0.103 0.046 -0.165 -0.250
(0.052) (0.026) (0.070) (0.041)
Unearned income/week -0.073 -0.009 0.124 0.128
(0.036) (0.040) (0.057) (0.042)
R2 0.248 0.231 0.404 0.386
*Also includes all three other lagged stress measures, a vector of measures of numbers and ages of children, year indicators and indicators of the respondent’s and spouse’s decadal ages (31-40 and 41+).
28
Table 5. Lag Structure of Stress in Response to the Addition of a Child*
Response in Standard-Deviation Units of Stress HUSBANDS WIVES
Years after birth:
Time stress
Financial stress
Time stress
Financial stress
HILDA 0-1 0.080 0.062 0.224 0.071 (0.028) (0.040) (0.035) (0.071)
1-2 0.032 0.054 0.119 0.028 (0.021) (0.049) (0.039) (0.046)
2-3 0.081 0.000 0.124 -0.029 (0.047) (0.068) (0.030) (0.071)
3-4 0.051 0.017 0.157 0.000 (0.070) (0.094) (0.064) (0.093)
SOEP
0-2 0.050 0.013 0.202 0.015 (0.049) (0.053) (0.055) (0.058)
2-4 -0.083 0.163 0.129 0.135 (0.117) (0.108) (0.149) (0.113)
*Based on LS coefficient estimates. Each underlying equation contains current values of all the regressors underlying the estimates in Tables 3a and 3b, except that it includes the lagged stress measures the year before the birth (two years in the SOEP). The equations for years after the initial year are restricted to couples who did not experience a second birth in the interval.
29
Table 6. Transfers/Extra Income Required to Reduce Wife's Financial Stress Equal to the Increase in Her Time Stress (in SD Units) from a Birth: Simulations from the HILDA and SOEP
Annual one-time cost per
new-parent household Annual cost per married couple
Simulation Description
HILDA ($)
SOEP (€)
HILDA ($)
SOEP (€)
1 Earnings transfer from husband to wife 144,788 296,607 14,189 23,729
2 Increase wife’s earnings 85,879 42,104 8,416 3,368
3 Increase husband’s earnings 211,075 49,069 20,685 3,926
30
Table 7. Year-to-Year Transition Matrices on Stress, with or without Child Departures, HILDA 2001-12, SOEP 2002-12
No departure: HILDA (N=23,869)
SOEP
(N=19,039)
HILDA (N=23,608)
SOEP (N=18,968)
Departure: (N=987) (N=1214) (N=987) (N=1214)
Time Stress
Change in Stress Men, No Departure Women, No Departure
Increase 20.8 26.4 21.4 27.5
Same 56.6 45.9 55.5 45.3
Decrease 22.6 27.7 23.1 27.2
Men, Departure Women, Departure
Increase 19.2 26.3 17.7 24.3
Same 57.2 45.3 57.5 46.7
Decrease 23.6 28.4 24.8 29.0
Financial Stress
Change in Stress Men, No Departure Women, No Departure
Increase 22.2 25.9 22.4 25.3
Same 54.0 50.4 52.9 50.1
Decrease 23.8 23.7 24.7 24.6
Men, Departure Women, Departure
Increase 18.8 25.2 21.2 24.9
Same 56.3 48.2 54.9 49.3
Decrease 24.9 26.6 23.9 25.8
31
Table 8. LS Estimates of Effects on Stress in Response to the Departure of a Child*
Response in Standard-Deviation Units of Stress HUSBANDS WIVES
Years after birth:
Time stress
Financial stress
Time stress
Financial stress
HILDA -0.038 -0.005 -0.057 0.042 (0.030) (0.040) (0.030) (0.038)
SOEP 0.0001 0.071 - 0.041 0.026 (0.039) (0.040) (0.035) (0.035)
*The underlying equations include all the variables in the specifications reported in Tables 3a and 3b, except that the vectors of indicators of respondents’ and spouses’ ages denote ages 41-50 and 50+.
32
Figure 1. Time and Financial Stress Before and After Birth of a Child, HILDA 2001-12
2.1
2.2
2.3
2.4
2.5
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
-4 -3 -2 -1 0 1 2 3 4
F
i
n
a
n
c
i
a
l
S
t
r
e
s
s
T
i
m
e
S
t
r
e
s
s
Years From Child's Birth
Men-Time Stress
Women-Time Stress
Men-Financial Stress
Women-Financial Stress
33
Figure 2. Time and Financial Stress Before and After Birth of a Child, SOEP 2002-12
2.4
2.5
2.6
2.7
2.8
2.6
2.7
2.8
2.9
3.0
3.1
3.2
3.3
3.4
-4 -2 0 2 4
F
i
n
a
n
c
i
a
l
S
t
r
e
s
s
T
i
m
e
S
t
r
e
s
s
Years From Child's Birth
Men-Time Stress
Women-Time Stress
Men-Financial Stress
Women-Financial Stress
34
Figure 3. Time and Financial Stress Before and After Departure of a Child, HILDA 2001-12
2.1
2.2
2.3
2.4
2.5
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
-4 -3 -2 -1 0 1 2 3 4
F
i
n
a
n
c
i
a
l
S
t
r
e
s
s
T
i
m
e
S
t
r
e
s
s
Years From Child's Departure
Men-Time Stress
Women-Time Stress
Men-Financial Stress
Women-Financial Stress
35
Figure 4. Time and Financial Stress Before and After Departure of a Child, SOEP 2002-12
2.4
2.5
2.6
2.7
2.8
2.6
2.7
2.8
2.9
3.0
3.1
3.2
3.3
3.4
-4 -2 0 2 4
F
i
n
a
n
c
i
a
l
S
t
r
e
s
s
T
i
m
e
S
t
r
e
s
s
Years From Child's Departure
Men-Time Stress
Women-Time Stress
Men-Financial Stress
Women-Financial Stress
36
DATA APPENDIX
A. Australian HILDA Survey
The sample comprises individuals who:
are married or in a de facto relationship in the current wave and the previous wave; are not in a same-sex relationship; are in a relationship with the woman between 18 and 45 years of age (inclusive); indicate they have the same partner in both waves (and both partners agree); live in the same household (with no other persons other than dependents); report valid responses for time stress and financial stress in the current and previous wave;
and report valid responses for them or their partner giving birth to (or adopting) a child in the
previous 12 months, and both partners agree.
Time stress is constructed from answers to the question, How often do you feel rushed or pressed for time?, which is asked in the self-completion part of the survey. Possible answers are: Almost always; Often; Sometimes; Rarely; and Never. The original values attached to these responses range from 1 to 5, respectively, but scores are reversed so that higher values represent higher stress levels.
Financial stress is the answer to the question, asked in the interview portion of the survey: I am now going to ask you some questions about how satisfied or dissatisfied you are with some of the things happening in your life. I am going to read out a list of different aspects of life and, using the scale on SHOWCARD [..], I want you to pick a number between 0 and 10 that indicates your level of satisfaction with each. The more satisfied you are, the higher the number you should pick. The less satisfied you are, the lower the number. The actual showcard shows a scale represented by a line with equally spaced ticks numbered 0 to 10 (from left to right). Only the two end points of the scale are labelled; 0 denotes ‘totally dissatisfied’ and 10 denotes ‘totally satisfied’ (10). The third entry on the list of eight aspects of life the respondent is asked to rate is Your financial situation?
Birth in past year is our measure of birth/adoption and uses information collected in the household relationships grid. This gives a precise indicator for a birth between two waves (on average 12 months apart).
Weekly hours paid employment plus commuting is based on the answers to the question, asked in the self-completion portion of the survey: How much time would you spend on each of the following activities in a typical week? Among the nine activities listed are Paid employment and Travelling to and from a place of paid employment. Respondents are instructed to make sure not to count any activity twice and if they do not spend time on a particular activity they record a zero. If either the paid employment or commuting component is missing, the sum (hours paid employment plus commuting) is also missing.
Weekly hours home production is based on the same question from which weekly hours of paid employment plus commuting is derived. The activities that make up home production are:
Household errands, such as shopping, banking, paying bills, and keeping financial records (but do not include driving children to school and to other activities).
Housework, such as preparing meals, washing dishes, cleaning house, washing clothes, ironing and sewing.
37
Outdoor tasks, including home maintenance (repairs, improvements, painting etc.), car maintenance or repairs and gardening.
Playing with your children, helping them with personal care, teaching, coaching or actively supervising them, or getting them to child care, school and other activities.
Looking after other people’s children (aged under 12years) on a regular, unpaid basis. Caring for a disabled spouse or disabled adult relative, or caring for elderly parents or
parents-in-law
If any of the six home production components is missing, the sum (weekly hours home production) is also missing.
Wages/Earnings is the derived variable ‘Current weekly gross wages and salary - all jobs ($) [imputed] [weighted topcode]’, which despite the label is actually the gross weekly in a usual week. It is the sum of wages and salary in the main job and other employment. Missing values for these components have been imputed (see Hayes and Watson, 2009). To preserve the weighted mean, top-coded variables have a value substituted which is the weighted average value of all cases which exceed the threshold.
Unearned income is constructed by taking household total gross income from all sources (excluding windfall income) for the preceding financial year and subtracting the component due to salaries and wages. Missing values for the components have been imputed (see Hayes and Watson, 2009). Note that this variable is a lagged variable by construction, although by how many months depends on when the respondent was interviewed (Australia’s financial year runs from 1 July to 30 June, whereas the bulk of respondents are interviewed between September and November each year).
Number of children are derived variables constructed from the household relationships grid. They represent the number of dependent children of particular ages in the household (indicated age ranges in the variable names are inclusive), and include partner’s children. In the event of a birth between waves, the number of children in the household aged 0 to 4 is reduced by 1 in the wave immediately following the birth only, to ensure the effect of the birth is picked up by the dedicated dummy variable ‘birth in last year’ and the new addition does not get double counted.
Very good health is based responses to the question: In general, would you say your health is … . The response options are: Excellent, Very good, Good, Fair, and Poor. We create a binary variable with responses to the first two categories indicating very good health.
B. German SOEP
The sample comprises individuals who:
are married or in a de-facto relationship in the current wave and two waves ago; are not in a same-sex relationship; are in a relationship and the woman is between 18 and 45 years of age inclusive; indicate they have the same partner in this wave as they did two waves ago (and both partners
agree); live in the same household (with no other persons other than dependents); report valid responses for time stress and financial stress in the current wave and two waves
ago; and
38
report valid responses for them or their partner giving birth to (or adopting) a child in the previous 12 months, and both partners agree
Time stress is based on responses to a question, asked in the individual questionnaire ‘Health and Illness’ section, that reads: Please think about the last four weeks. How often did it occur within this period of time, that [...]. It then asks about 8 specific domains, one of them which reads ... you felt rushed or pressed for time? Possible answers to this question are: Always, Often, Sometimes, Hardly ever, and Never. These are recoded from 1 to 5 with higher levels representing greater stress.
Financial stress is derived from answers to the question, asked in the individual questionnaire ‘Your current life situation’ section: How satisfied are you today with the following areas of your life? Please answer by using the following scale: 0 means "totally unhappy", 10 means "totally happy". How satisfied are you with [...]. It then asks about 11 specific domains, one of which reads your household income? The actual questionnaire shows a scale of equally spaced blocks numbered 0 to 10 (from left to right). Only the two end points of the scale are labelled; 0 denotes ‘totally unhappy’ and 10 denotes ‘totally happy’.
Birth in last two years is based on the penultimate question ‘Family situation and background’ section in the individual questionnaire, which asks about 12 specific life events related to family. Specifically, the question reads: Has your family situation changed since [wave-specific date]? Please indicate if any of the following apply to you and if so, when this change occurred. One of the 12 family related life events is Had a child. Due to the interaction of the wave-specific date in the question itself, the date of birth, and the date of the actual interview it is not straightforward to get an indicator for a birth between this and the previous wave, but very easy to establish if a birth occurred between the current wave and two waves ago by using the variable fnpar0593.
Weekly hours paid employment is from the individual questionnaire ‘Your current employment’ section, respondents are first asked about their contractual working hours (if they are employed), followed by the question (which forms the basis of our variable): And how many hours do your actual working-hours consist of including possible over-time? with responses required for a week. Unlike the HILDA Survey data, this measure does not include commuting time.
Hours home production is based on a series of variables related to time use. In the section ‘Your current life situation’ of the individual questionnaire (which also collects the information on satisfaction with household finances) respondents are asked: What does a typical weekday look like for you? How many hours per day do you spend on the following activities? Please give only whole hours. Use zero if the activity does not apply! It then asks about eight specific activities. We construct household production by summing the amount of time spent on: supporting persons in care; running errands; doing housework; caring for children; and doing repairs around the house. If any of the home production components is missing, the sum (weekly hours home production) is also missing.
Wages/Earnings is based on (generated) current gross labour income per month in Euro. The amount is divided by 1000 x 4.3 to obtain weekly amounts in EUR to correspond with the weekly amounts in AUD for the HILDA Survey data. The underlying question on which the variable is based comes from the individual questionnaire ‘Your current employment’ section and reads: How high was your income from employment last month? If you received extra income such as vacation pay or back pay, please do not include this. Please do include overtime pay. If you are self-employed: Please estimate your monthly income before and after tax. Please fill in both: gross income, which means wages or salary before deduction of taxes and social security; and net income, which means the sum after
39
deduction of taxes, social security, and unemployment and health insurance. If the information was missing the data was imputed by the SOEP team (hence the reference to ‘generated’ in the variable description).
Unearned income is constructed by taking the household’s pre-Government income and subtracting household labour income, both of which are annual amounts and apply to the previous calendar year. Household public transfers and Social Security pensions are then added in. This amount is divided by 52 (weeks) and 1000 to obtain a measure expressed in Euros per week. Pre-Government income is the sum of total family income from labor earnings, asset flows, private retirement income and private transfers. Labor earnings include wages and salary from all employment including training, self-employment income, and bonuses, overtime, and profit-sharing. Asset flows include income from interest, dividends, and rent. Private transfers include payments from individuals outside of the household including alimony and child support payments. In order to arrive at unearned income, the component from labor earnings is then subtracted. Specifically, labor earnings is the sum of income from the primary job, any secondary jobs, self-employment, service pay, 13th month pay, 14th month pay, Christmas bonus pay, holiday bonus pay, miscellaneous bonus pay, profit-sharing income, indemnity payments, and commuting expenses or travel grants.
Numbers of children are based on the information in the household question form. In case there has been a birth in between the last two waves (because of information on time and financial stress is collected biannually), the number of children aged 0 to 4 is reduced by 1 in the wave following the birth only, to ensure the effect of the birth is picked up by the dedicated dummy variable ‘birth in last two years’ and the new addition does not get double counted.
Very good health is based on responses to the question: How would you describe your current health? Possible responses are: Very good, Good, Satisfactory, Poor, and Bad. We create a binary variable with responses to the first two categories indicating very good health.